1 / 43

Automatic Speech Recognition: An Overview

Automatic Speech Recognition: An Overview. Julia Hirschberg CS 4706 (Thanks to Roberto Pieraccini and Francis Ganong for some slides). DIALOG. SEMANTICS. SPEECH RECOGNITION. SPOKEN LANGUAGE UNDERSTANDING. SYNTAX. LEXICON. MORPHOLOGY. SPEECH SYNTHESIS. PHONETICS. DIALOG

abba
Télécharger la présentation

Automatic Speech Recognition: An Overview

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Automatic Speech Recognition: An Overview Julia Hirschberg CS 4706 (Thanks to Roberto Pieraccini and Francis Ganongfor some slides)

  2. DIALOG SEMANTICS SPEECH RECOGNITION SPOKEN LANGUAGE UNDERSTANDING SYNTAX LEXICON MORPHOLOGY SPEECH SYNTHESIS PHONETICS DIALOG MANAGEMENT INNER EAR ACOUSTIC NERVE VOCAL-TRACT ARTICULATORS Recreating the Speech Chain

  3. NP NP VP SEVEN THREE ZERO IS MY FOUR NUMBER TWO NINE SEVEN & E E & & r n b n n th ü e n o i s v z O & r r f m n t I e m v s r I The Problem of Segmentation... or... Why Speech Recognition is so Difficult (user:Roberto (attribute:telephone-num value:7360474))

  4. Ellipses and Anaphors Limited vocabulary Multiple Interpretations Speaker Dependency Word variations NP NP VP Word confusability SEVEN THREE ZERO IS MY Context-dependency FOUR NUMBER TWO NINE SEVEN Coarticulation Noise/reverberation E & E & & r n n o th e ü n n b i s v O z & r r m e f n m I t s v r I Intra-speaker variability The Illusion of Segmentation... or... Why Speech Recognition is so Difficult (user:Roberto (attribute:telephone-num value:7360474))

  5. Speech Recognition: the Early Years • 1952 – Automatic Digit Recognition (AUDREY) • Davis, Biddulph, Balashek (Bell Laboratories)

  6. 1960’s – Speech Processing and Digital Computers • AD/DA converters and digital computers start appearing in the labs James Flanagan Bell Laboratories

  7. J. R. Pierce Executive Director, Bell Laboratories 1969 – Whither Speech Recognition? General purpose speech recognition seems far away. Social-purpose speech recognition is severely limited. It would seem appropriate for people to ask themselves why they are working in the field and what they can expect to accomplish… It would be too simple to say that work in speech recognition is carried out simply because one can get money for it. That is a necessary but not sufficient condition. We are safe in asserting that speech recognition is attractive to money. The attraction is perhaps similar to the attraction of schemes for turning water into gasoline, extracting gold from the sea, curing cancer, or going to the moon. One doesn’t attract thoughtlessly given dollars by means of schemes for cutting the cost of soap by 10%. To sell suckers, one uses deceit and offers glamour… Most recognizers behave, not like scientists, but like mad inventors or untrustworthy engineers. The typical recognizer gets it into his head that he can solve “the problem.” The basis for this is either individual inspiration (the “mad inventor” source of knowledge) or acceptance of untested rules, schemes, or information (the untrustworthy engineer approach). The Journal of the Acoustical Society of America, June 1969

  8. 1971-1976: The ARPA SUR project • Despite anti-speech recognition campaign led by Pierce Commission ARPA launches 5 year Spoken Understanding Research program • Goal: 1000-word vocabulary, 90% understanding rate, near real time on 100 mips machine • 4 Systems built by the end of the program • SDC (24%) • BBN’s HWIM (44%) • CMU’s Hearsay II (74%) • CMU’s HARPY (95% -- but 80 times real time!) • Rule-based systems except for Harpy • Engineering approach: search network of all the possible utterances LESSON LEARNED: Hand-built knowledge does not scale up Need of a global “optimization” criterion Raj Reddy -- CMU

  9. Lack of clear evaluation criteria • ARPA felt systems had failed • Project not extended • Speech Understanding: too early for its time • Need a standard evaluation method so that progress within and across systems could be measured

  10. Isolated Words Speaker Dependent Connected Words Speaker Independent Sub-Word Units 1970’s – Dynamic Time WarpingThe Brute Force of the Engineering Approach T.K. Vyntsyuk (1968) H. Sakoe, S. Chiba (1970) TEMPLATE (WORD 7) UNKNOWN WORD

  11. Fred Jelinek Acoustic HMMs Word Tri-grams a11 a22 a33 a12 a23 S1 S2 S3 1980s -- The Statistical Approach • Based on work on Hidden Markov Models done by Leonard Baum at IDA, Princeton in the late 1960s • Purely statistical approach pursued by Fred Jelinek and Jim Baker, IBM T.J.Watson Research Jim Baker • No Data Like More Data • “Whenever I fire a linguist, our system performance improves” (1988) • Some of my best friends are linguists (2004)

  12. 1980-1990 – Statistical approach becomes ubiquitous • Lawrence Rabiner, A Tutorial on Hidden Markov Models and Selected Applications in Speech Recognition, Proceeding of the IEEE, Vol. 77, No. 2, February 1989.

  13. 1995 1997 1996 1998 1999 2000 2001 HOSTING 2002 2003 MIT 2004 … APPLICATION DEVELOPERS STANDARDS TOOLS SRI PLATFORM INTEGRATORS STANDARDS TECHNOLOGY VENDORS STANDARDS 1980s-1990s – The Power of Evaluation SPOKEN DIALOG INDUSTRY SPEECHWORKS NUANCE Pros and Cons of DARPA programs + Continuous incremental improvement - Loss of “bio-diversity”

  14. NUANCE Today

  15. Large Vocabulary Continuous Speech Recognition (LVCSR) Today ~20,000-64,000 words Speaker independent (vs. speaker-dependent) Continuous speech (vs isolated-word) Conversational speech, dialogue (vs. dictation) Many languages (vs. English, Japanese, Swedish, French) 3/21/2012 15 Speech and Language Processing Jurafsky and Martin

  16. Current error rates 3/21/2012 16 Speech and Language Processing Jurafsky and Martin

  17. Human vs. Machine Error Rates • Conclusions: • Machines about 5 times worse than humans • Gap increases with noisy speech • These numbers are rough… 3/21/2012 17 Speech and Language Processing Jurafsky and Martin

  18. How to Build an ASR System • Build a statistical model of the speech-to-text process • Collect lots of speech and transcribe all the words • Train the acoustic model on the labeled speech • Train a language model on the transcription + lots more text • Paradigms: • Supervised Machine Learning + Search • The Noisy Channel Model

  19. ASR Paradigm • Given an acoustic observation: • What is the most likely sequence of words to explain the input? • Using an • Acoustic Model and a • Language Model • Two problems: • How to score hypotheses (Modeling) • How to pick hypotheses to score (Search)

  20. The Noisy Channel Model • Search through space of all possible sentences. • Pick the one that is most probable given the waveform • Decoding

  21. The Noisy Channel Model: Assumptions • What is the most likely sentence out of all sentences in the language L, given some acoustic input O? • Treat acoustic input O as sequence of individual acoustic observations or frames • O = o1,o2,o3,…,ot • Define a sentence W as a sequence of words: • W = w1,w2,w3,…,wn

  22. Noisy Channel Model • Probabilistic implication: Pick the highest probable sequence: • We can use Bayes rule to rewrite this: • Since denominator is the same for each candidate sentence W, we can ignore it for the argmax:

  23. Speech Recognition Meets Noisy Channel: Acoustic Likelihoods and LM Priors

  24. Components of an ASR System • Corpora for training and testing of components • Representation for input and method of extracting features • Pronunciation Model (a lexicon) • Acoustic Model (acoustic characteristics of subword units) • Language Model (word sequence probabilities) • Algorithms to search hypothesis space efficiently

  25. Training and Test Corpora • Collect corpora appropriate for recognition task at hand • Small speech + phonetic transcription to associate sounds with symbols (Acoustic Model) • Large (>= 60 hrs) speech + orthographic transcription to associate words with sounds (Acoustic Model+) • Very large text corpus to identify ngram probabilities or build a grammar (Language Model)

  26. Building the Acoustic Model • Goal: Model likelihood of sounds given spectral features, pronunciation models, and prior context • Usually represented as Hidden Markov Model • States represent phones or other subword units for each word in the lexicon • Transition probabilities (a)on states: how likely to transition from one unit to itself? To the next? • Observation likelihoods (b): how likely is spectral feature vector (the acoustic information) to be observed at state i?

  27. Training a Word HMM

  28. Initial estimates of acoustic models from phonetically transcribed corpus or flat start • Transition probabilities between phone states • Observation probabilities associating phone states with acoustic features of windows of waveform • Embedded training: • Re-estimate probabilities using initial phone HMMs + orthographically transcribed corpus + pronunciation lexicon to create whole sentence HMMs for each sentence in training corpus • Iteratively retrain transition and observation probabilities by running the training data through the model until convergence

  29. Training the Acoustic Model Iteratively sum over all possible segmentations of words and phones – given the transcript -- re-estimating HMM parameters accordingly until convergence

  30. Building the Pronunciation Model • Models likelihood of word given network of candidate phone hypotheses • Multiple pronunciations for each word • May be weighted automaton or simple dictionary • Words come from all corpora (including text) • Pronunciations come from pronouncing dictionary or TTS system

  31. ASR Lexicon: Markov Models for Pronunciation

  32. Building the Language Model • Models likelihood of word given previous word(s) • Ngram models: • Build the LM by calculating bigram or trigram probabilities from text training corpus: how likely is one word to follow another? To follow the two previous words? • Smoothing issues: sparse data • Grammars • Finite state grammar or Context Free Grammar (CFG) or semantic grammar • Out of Vocabulary (OOV) problem

  33. Search/Decoding • Find the best hypothesis Ŵ (P(O|W) P(W)) given • A sequence of acoustic feature vectors (O) • A trained HMM (Acoustic Model) • Lexicon (Pronunciation Model) • Probabilities of word sequences (Language Model) • For O • Calculate most likely state sequences in HMM given transition and observation probabilities  lattice w/AM probabilities • Trace backward thru lattice to assign words to states using AM and LM probabilities (decoding) • N best vs. 1-best vs. lattice output • Limiting search • Lattice minimization and determinization • Pruning: beam search

  34. Evaluating Success • Transcription • Goal: Low Word Error Rate (Subst+Ins+Del)/N * 100 • This is a test Thesis test. (1subst+2del)/4*100=75% WER That was the dentist calling. (4 subst+1ins)/4words * 100=125% WER • Understanding • Goal: High Concept Accuracy • How many domain concepts were correctly recognized? I want to go from Boston to Washington on September 29

  35. Domain concepts Values • source city Boston • target city Washington • travel monthSeptember • travel day 29 • Ref: Go from Boston to Washington on September 29vs. ASR: Go to Boston from Washington on December 29 • 3concepts/4concepts * 100 = 25% Concept Error Rate or 75% Concept Accuracy • 3subst/8words * 100 = 37.5% WER or 62.5% Word Accuracy • Which is better?

  36. ASR Today • Combines many probabilistic phenomena: varying acoustic features of phones, likely pronunciations of words, likely sequences of words • Relies upon many approximate techniques to ‘translate’ a signal • Finite State Transducers now a standard representation for all components

  37. ASR Future • 3->5 • Triphones -> Quinphones • Trigrams -> Pentagrams • Bigger acoustic models • More parameters • More mixtures • Bigger lexicons • 65k -> 256k

  38. Larger language models • More data, more parameters • Better back-offs • More kinds of adaptation • Feature space adaptation • Discriminative training instead of MLE • Rover: combinations of recognizers • FSM flattening of knowledge into uniform structure

  39. But What is Human about State-of-the-Art ASR? Input Wave Acoustic Features Acoustic Models Front End Search Lexicon Language Models

  40. Acoustic Features Input Wave Acoustic Features Acoustic Models Front End Front End: MFCC Search Postprocessing Language Models Lexicon Input Wave Sampling, Windowing FastFourierTransform cosine transform first 8-12 coefficients Mel Filter Bank: Stacking, computation of deltas:Normalizations: filtering, etc Linear Transformations:dimensionality reduction

  41. Basic Lexicon Input Wave Acoustic Features Acoustic Models Front End Search • A list of spellings and pronunciations • Canonical pronunciations • And a few others • Limited to 64k entries • Support simple stems and suffixes • Linguistically naïve • No phonological rewrites • Doesn’t support all languages Language Models Lexicon

  42. Language Model • Frequency sensitive, like humans • Context sensitive, like humans

  43. Next Class • Building an ASR system: the Pocket Sphinx Toolkit

More Related